Junior AI Interpretability Engineer (Remote)

University of Chicago; Chicago (IL); Hybrid; USD 100,000 - 130,000; Full time

Job summary

University of Chicago seeks a researcher and software developer to lead day-to-day efforts in AI interpretability for scientific discovery within Globus Labs. You will design and implement GPU-accelerated workflows, apply mechanistic interpretability to scientific foundation models, and collaborate with researchers across materials science, chemistry, and x-ray science.

Ideal candidates have strong Python, PyTorch/JAX experience, and a track record in building reusable data pipelines and

Qualifications

• Experience developing Python code and scientific computing workflows.
• Familiarity with deep learning frameworks (PyTorch/JAX) and Transformer models.
• Understanding of mechanistic interpretability methods and scientific data domains.

Responsibilities

• Coordinate day-to-day activities of interpretability for scientific discovery.
• Develop tooling, pipelines, and GPU-accelerated workflows for experiments.
• Assist in defining research roadmap and align with lab goals.
• Collaborate across teams to integrate features and fix issues.
• Support external engagement and partnerships with program managers.

Skills

Python

Node.js

C/C++

PyTorch

JAX

Transformers

Git

REST APIs

CLI tools

NumPy

pandas

scikit-learn

Education

Bachelor's degree in related field

Tools

Git

Slurm

AWS/GCP/Azure

REST APIs

Job description

University of Chicago seeks a researcher and software developer to lead day-to-day efforts in AI interpretability for scientific discovery within Globus Labs. You will design and implement GPU-accelerated workflows, apply mechanistic interpretability to scientific foundation models, and collaborate with researchers across materials science, chemistry, and x-ray science.

Ideal candidates have strong Python, PyTorch/JAX experience, and a track record in building reusable data pipelines and